architectures · bigcode
Code-generation transformer using grouped-query attention and sliding-window attention, trained with fill-in-the-middle so the model can condition on text after the cursor as well as before it. Published by the BigCode open-governance project with its training corpus (The Stack v2) documented alongside. Declared as model_type `starcoder2`.
| Publisher | bigcode |
|---|---|
| Origin | International |
| Licence | BigCode-OpenRAIL-M |
Registry record as of 2026-08-14. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.